What happened
Springer Nature published updated editorial policies on artificial intelligence on 10 September 2026, together with guidance written separately for authors, peer reviewers and editors (Springer Nature, 2026). The publisher presented the update as a response to growing demand from the research community for clear and practical direction on using AI across the publishing process. Trade coverage reported that the publisher cited research in which more than 70% of researchers said publisher guidance on AI was becoming more important to them (Knowledgespeak, 2026).
The main change is one of approach. Earlier rules across the sector largely asked a yes or no question: did you use a generative AI tool, and did you declare it? The new framework instead asks how AI was used, what impact that use could have, how much risk it introduces and whether appropriate human oversight was maintained (Springer Nature, 2026). The publisher says the policies will be reviewed periodically as technology, regulation and community expectations change (Springer Nature, 2026).
What the new framework says
The policy separates uses by risk. Lower-risk uses that support efficiency, clarity or accessibility, such as improving language or cleaning data, sit at one end. Higher-risk uses that can shape interpretation or decisions, such as identifying patterns in data or interpreting results, need more scrutiny. Uses that replace accountable human judgement or compromise publishing integrity, such as generating conclusions or delegating peer review to AI, are not allowed (Springer Nature, 2026; Knowledgespeak, 2026).
The accompanying guidance keeps several long-standing rules. AI tools cannot be named as authors because they cannot be held accountable; generative AI use should be declared in the manuscript, although AI-assisted copy editing does not need a declaration; and AI-generated images are generally not permitted outside a limited set of exceptions (Springer Nature, n.d.). For reviewers, the guidance stresses confidentiality: manuscript content should not be uploaded to unsecured or public AI systems, and AI may support but not replace a reviewer's own expertise and judgement (Springer Nature, n.d.).
Background: a fast-moving and uneven landscape
Publishers first set ground rules soon after ChatGPT appeared. In January 2023 Nature stated that no large language model would be accepted as a credited author and that any use should be documented (Nature, 2023), and COPE took the same position on authorship, stressing that human authors remain responsible for all content (COPE, 2023). Guidance has since spread but remains inconsistent: a bibliometric study of the 100 largest publishers and 100 highly ranked journals found that only 24% of the publishers but 87% of the journals provided guidance, and that some journal rules conflicted with those of their own publishers (Ganjavi et al., 2024).
Peer review is a particular concern. An analysis of reviews submitted to major AI conferences estimated that between 6.5% and 16.9% of review text could have been substantially modified by large language models (Liang et al., 2024). Publisher positions on reviewer use still differ widely: a September 2026 survey of policies found that some publishers permit limited, disclosed use while others forbid reviewers from entering any part of a manuscript into AI tools (Weinberg, 2026).
Why it matters for PhD students
Doctoral researchers increasingly use AI tools for drafting support, coding and literature searching, and many review papers for the first time during the PhD. A risk-based policy rewards being able to explain exactly what a tool did and how you checked it. Polishing grammar is unlikely to raise concern; using a tool to find patterns or suggest interpretations places the burden on you to show human oversight (Springer Nature, 2026).
Policies also differ between publishers and sometimes between journals within one publisher (Ganjavi et al., 2024; Weinberg, 2026). The rules that apply to a manuscript are those of the specific journal you submit to, and the rules for a review invitation are those of the journal that sent it. Your university and funder may add their own requirements on top.
